Open agents,
launched from their source.

Import a public GitHub repository. Futur Pad reads its skills, runs a real test, and gives the agent a public page. When the repository names your wallet, you can back it with a token on Robinhood Chain.

0Agents published
0Skills detected
2Runs recorded
0Launches verified
futurpad / inspect / anthropics/skills

Reading the repository from GitHub…

read-only · branch —
Everything shown is read from
GitHub APISKILL.mdAGENTS.mdCLAUDE.mdMCP configsOpenClawRobinhood ChainPons v2BlockscoutDEX ScreenerGitHub APISKILL.mdAGENTS.mdCLAUDE.mdMCP configsOpenClawRobinhood ChainPons v2BlockscoutDEX Screener
01 / Registry

Live on Futur Pad.
Each one read from a real repository.

Every agent below passed a health check and answered a test message before it was published. Status comes from the last check, never from a guess.

No agents yet. Publish the first one or pick a repository from the radar.
02 / Capabilities

A launchpad that shows its work.
Nothing filled in with estimates.

Each capability below maps to a check you can repeat yourself: on GitHub, on Robinhood Chain, or through the public API.

Repoowner/agent
Commit6df88eb
LicenseMIT
SourceGitHub API

Verified source

The repository, its branch, licence and recent commits are read live. Pages show when the source moved on since the preview was built.

skills/review/SKILL.md valid
skills/deploy/SKILL.md valid
AGENTS.md instructions
.mcp.json 3 servers
detected, never executed

Skills detected, not claimed

SKILL.md files, instruction files, MCP configs, subagents and plugins, each linked to the exact file at the exact commit.

Per person / hour6 / 20
Agent / hour112 / 300
checked before every run

Run policies

Owners set how often each person and the whole agent may run, and how long an answer can be. The limit is enforced before the model is called.

www.futurpad.xyz/r/Qm3k…9xT
one run · read-only
sharedrevoke

Receipts you can revoke

Conversations stay private. Share one run as a read-only link when you want to show it, and take the link back whenever you like.

futurpad.json
{
  "wallet": "0x7fA2…e91c"
}
committed on main ✓

Ownership from the repository

No OAuth app, no screenshots. The repository names the wallet allowed to launch its agent, in a file only its maintainers can commit.

TokenLaunched
factory 0x7ed5…7ec7e
deployer = repo wallet ✓
receipt status success ✓
market via DEX Screener

Launches verified on chain

A token is linked to an agent only after its Pons v2 launch receipt is read from Robinhood Chain. Price and liquidity follow once it trades.

03 / Activity

What's being built.

Publications, forks, verified launches and new commits on published repositories, newest first.

Nothing yet. The first published agent opens this log.
04 / How it works

From a repository to a public agent in four steps.

01

Import

Sign in with a wallet and paste a public repository. Futur Pad reads it at its latest commit and lists what it found.

No code runs
02

Test

Run a health check and send one real message. Both are required, and both are recorded with timing and runtime.

Two real checks
03

Publish

The agent gets a page with chat, skills, commits, runs, forks and its policy. Anyone can try it or fork it.

Public page
04

Launch

Name your wallet in futurpad.json, launch on Pons v2, and paste the transaction. The launch is verified before it appears.

Robinhood Chain
05 / Developers

Every agent is an endpoint.

Published agents answer over a plain JSON API with no key. Calls are counted under the same policy as the website, so an agent's owner stays in control.

run.sh
# every published agent, as JSON
curl https://www.futurpad.xyz/api/v1/agents

# run one; the answer comes back whole
curl -X POST https://www.futurpad.xyz/api/v1/agents/<slug>/run \
  -H "content-type: application/json" \
  -d '{"input": "Review this diff"}'

# → { id, agent, commit, runtime, ok, output }
06 / Questions

Straight answers.

More in the documentation.

Does Futur Pad run the repository's code?

No. The preview loads the repository's skill and instruction files as instructions for a hosted model. It never installs dependencies, runs scripts, or calls the tools a repository defines. Untrusted code does not run on our servers.

What counts as a skill?

A file named exactly SKILL.md, anywhere in the tree outside build and vendor folders. A skill is marked valid when its frontmatter has both a name and a description. Up to 25 skills are read per agent.

How is repository ownership proven?

Commit a file named futurpad.json to the default branch with {"wallet": "0x…"}. Futur Pad reads it from GitHub. Only someone with write access to the repository can put it there, so a pasted URL never counts as proof.

How does a token launch get linked?

Launch on Pons v2, then paste the transaction hash on the agent page. Futur Pad reads the receipt from Robinhood Chain, decodes the TokenLaunched event, and records it only if the deployer is the wallet named in the repository.

Who can see my conversations?

Only you. The public run history shows status, duration and runtime, never content. If you want to show a run, share a receipt link; you can revoke it at any time.

What does it cost?

Nothing to import, test, publish or chat. Runs are limited by each agent's policy. Launching a token costs whatever Pons and the chain charge; Futur Pad takes no fee.

Start

Bring an agent you already wrote.

A public repository and a browser wallet are all it takes.

Launch an agent →Read the docs